The SLUG Zinc-Finger Protein Represses E-Cadherin in Breast Cancer¹

E-cadherin is a cell-cell adhesion molecule that participates in homotypic, calcium-dependent interactions to form the epithelial adherens junction. This function is critical in the development and maintenance of a polar epithelium. Loss of E-cadherin expression has been demonstrated in carcinomas arising in many tissues, and E-cadherin loss is believed to contribute to both cancer development and progression (1). Work in our laboratory, based on the analysis of somatic cell hybrids, demonstrated the loss of E-cadherin in breast cancer cell lines was because of dominant transcriptional defects, suggesting the presence of a transcriptional repression pathway that extinguishes E-cadherin expression (2). We narrowed the critical region of the E-cadherin promoter responsible for E-cadherin expression defects in breast cancer to bp −108 to +125 (2). Other groups have also implicated the proximal E-cadherin promoter and E-box elements contained therein in transcriptional inactivation of Ecadherin in some cancer cell lines (3, 4, 5). Recent studies of the E-cadherin promoter have additionally supported the critical role of the proximal promoter region in regulating E-cadherin expression. The work has also highlighted specific transcription factors that may function in repression of E-cadherin in cancer. These factors include the zinc-finger transcription factors SNAIL (4, 6), δEF1/ZEB-1 (7), and SIP1/ZEB-2 (5), and the basic helix-loop-helix factor E12/E47 (8). The specific factors that repress E-cadherin likely vary depending on cell type and context. Additionally, the various E-cadherin repression factors described to date may act alone or in concert, and there may be other currently undefined factors required for transcriptional silencing of E-cadherin in cancer cells. The goal of the studies described here was to determine the specific promoter element(s) and factor(s) critical for repression of E-cadherin in breast carcinomas. We found both SNAIL and its family member SLUG to be capable of repressing E-cadherin in epithelial cells via the E-box elements in the proximal E-cadherin promoter. However, SLUG expression showed a much stronger correlation with loss of E-cadherin in breast cancer cell lines than did SNAIL expression, suggesting SLUG is a likely in vivo repressor of E-cadherin expression in breast carcinoma.

The RK3E cell line was provided by J. M. Rupert (University of Alabama, Birmingham, AL). The amphotropic Phoenix retrovirus packaging line was provided by G. Nolan (Stanford University, Stanford, CA). Both RK3E cells and Phoenix cells were grown in DMEM (Invitrogen Corp., Carlsbad, CA) supplemented with 10% fetal bovine serum (Invitrogen Corp.) and 1% penicillin/streptomycin (Invitrogen Corp.). All of the other cell lines were acquired from the American Type Culture Collection (Manassas, VA) and maintained as recommended. For the generation of stable RK3E clonal lines expressing ER³ chimera constructs, selection was carried out in medium supplemented with 1 μg/ml puromycin (Sigma Chemical Co., St. Louis, MO). After selection, the cell lines were maintained in medium containing 0.5 μg/ml puromycin. Designated cell lines were treated with medium supplemented with 0.5 μm 4-OHT (Sigma Chemical Co.) dissolved in 100% ethanol or mock treated with medium supplemented with 100% ethanol for the indicated time periods. SLUG- and SNAIL-expressing retroviruses were generated by transfecting plasmids into Phoenix packaging cells using FuGene6 as per the manufacturer’s instructions (Roche Molecular Biochemicals, Indianapolis, IN). After transfection (24 h) into the packaging line, the medium was changed and an additional 24 h later virus-containing supernatant was harvested. The supernatant was filtered and diluted 1:1 with fresh medium. It was then supplemented with 4 μg/ml Polybrene (Sigma Chemical Co.) and used to infect MDA-MB-468 cells. After infection of MDA-MB-468 (48 h), selection was initiated in 0.5 mg/ml Geneticin (Invitrogen Corp.). Total protein lysates were prepared 2 weeks after commencing drug selection.

Luciferase reporter gene constructs containing wild-type Ecadherin promoter sequences were described in detail previously (2, 9). Briefly, E-cadherin promoter sequences were amplified by PCR and cloned into the pGL2-Basic vector (Promega Corp., Madison, WI) upstream of firefly Luciferase. In all of the reporter gene constructs, the endogenous initiating methionine of the E-cadherin gene, located at bp +125, has been destroyed, and an additional 33 bp of flanking sequence separate E-cadherin promoter sequences from the Luciferase initiating methionine. PCR-based site-directed mutagenesis was used for the generation of reporter gene constructs with E-box mutations. All of the mutant constructs were made within the context of the reporter gene construct Ecad(-108)-Luc, which contains E-cadherin promoter sequences from −108 to +125 of the endogenous E-cadherin gene upstream of firefly Luciferase. E-box elements were mutated from 5′-CANNTG-3′ to 5′-AANNTA-3′ (sense strand). Full-length cDNAs for human SLUG and SNAIL were amplified from cell line RNA by reverse transcription-PCR, and a COOH-terminal flag epitope tag was added by PCR. Constructs were subcloned into the retroviral expression vector pPGS-CMV-CITE-neo (gift of G. Nabel, NIH, Bethesda, MD). For the generation of vectors encoding SLUG- and SNAIL-ER fusion proteins, SLUG and SNAIL cDNAs were cloned into the pBabePuro plasmid (gift of A. Friedman, Johns Hopkins University, Baltimore, MD) upstream of a modified mouse ERα ligand-binding domain. The identities of all plasmid inserts and vector boundary regions were confirmed by sequence analysis. The pCH110 plasmid contains a functional lacZ gene expressed under the control of the SV40 early promoter (Amersham Biosciences, Piscataway, NJ).

Cell lines growing at ∼70% confluence in six-well plates were transfected using FuGene6 (Roche Molecular Biochemicals) according to the manufacturer’s protocol. For experiments assessing activation of E-cadherin reporter gene constructs by endogenous factors, 0.8 μg E-cadherin reporter gene construct and 0.8 μg pCH110 were transfected per well. For experiments to determine the effects of SLUG and SNAIL on E-cadherin reporter gene activity, 1 μg effector plasmid (empty expression vector, SLUG expression vector, or SNAIL expression vector), 0.5 μg E-cadherin reporter gene construct, and 0.5 μg pCH110 were transfected per well. Experiments on the dose-dependent repression of E-cadherin reporter gene constructs by SLUG and SNAIL used 0.15 μg reporter construct, 0.35 μg pCH110, and increasing amounts of effector plasmid. In these dose-response studies, the total amount of transfected DNA was kept constant by adding empty expression vector as necessary. Cell extracts were prepared 36–40 h after transfection using reporter lysis buffer (Promega Corp.) followed by determination of luciferase and β-galactosidase activities. β-Galactosidase activity was used to normalize for transfection efficiency.

Total RNA was isolated from cells using TRIzol reagent (Invitrogen Corp.). Electrophoretic separation and membrane transfer of RNA were carried out by standard methods. For use as probes, E-cadherin, SLUG, SNAIL, and GAPDH cDNA fragments were labeled with [³²P]dCTP (Amersham Biosciences) by random priming with the RediPrime II kit (Amersham Biosciences). Prehybridization and hybridization were carried out in Rapid-Hyb buffer (Amersham Biosciences), the membrane was washed, and the blot was exposed to BioMax MS film (Kodak, Rochester, NY).

GST fusion proteins were generated by subcloning cDNA sequences corresponding to the amino half of either SLUG (amino acids 1–151) or SNAIL (amino acids 1–146) into the vector pGEX-2T (Amersham Biosciences). Plasmids were introduced into the Escherichia coli strain BL21, and a large-scale preparation of recombinant protein was performed. After sonication of the bacteria, the fusion protein-containing supernatant was collected by centrifugation. GST fusion proteins were purified from this supernatant on a glutathione Sepharose 4B (Amersham Biosciences) column. Purified recombinant GST-SLUG protein was used directly as antigen for antibody production. Purified recombinant GST-SNAIL protein was separated by electrophoresis on a SDS-polyacrylamide gel, and the band corresponding to full-length GST-SNAIL was excised and used as antigen. Rabbit injection and serum collection were carried out by Covance Research Products Inc. (Richmond, CA). Serum was ammonium sulfate precipitated and then used either directly (anti-SNAIL antibodies) or purified (anti-SLUG antibodies) against a recombinant maltose-binding protein (MBP) fusion protein using the AminoLink Plus Immobilization kit (Pierce, Rockford, IL). Recombinant MBP-SLUG was generated by subcloning cDNA sequences corresponding to the amino half of SLUG (amino acids 1–151) into the plasmid pMAL-c2 (New England Biolabs, Inc., Beverly, MA), inducing recombinant protein expression and purifying the MBP-SLUG on a column of amylose resin (New England Biolabs, Inc.).

Whole cell lysates were prepared in radioimmunoprecipitation assay buffer [150 mm NaCl, 0.5% deoxycholic acid, 0.1% SDS, 1% NP40, 50 mm Tris (pH 8.0)] with Complete protease inhibitors (Roche Molecular Biochemicals). Approximately 35 μg of total protein per sample were separated by electrophoresis on SDS-polyacrylamide gels and transferred to Immobilon P membranes (Millipore Corp., Bedford, MA) by semi-dry electroblotting (Transblot; Bio-Rad Laboratories, Hercules, CA). Primary antibodies and the dilutions used were as follows: mouse monoclonal anti-flag M2 (Sigma Chemical Co.), 1:5,000; mouse monoclonal antibody HECD-1 against E-cadherin (Zymed Laboratories, Inc., San Francisco, CA), 1:5,000; rabbit polyclonal antibody A2066 against β-actin (Sigma Chemical Co.), 1:1,000; rabbit polyclonal anti-SNAIL, 1:2,000; and rabbit polyclonal anti-SLUG, 1:800. Secondary antibodies and the dilutions used were as follows: horseradish peroxidase-conjugated goat antimouse IgG antibody (Pierce), 1:20,000; horseradish peroxidase-conjugated donkey antirabbit IgG antibody (Pierce); 1:20,000 for antiactin blots; and 1:80,000 for anti-SNAIL and anti-SLUG blots. Antibody complexes were detected using Western Blot Chemiluminescence Reagent Plus (NEN Life Science Products, Boston, MA) followed by exposure to X-OMAT AR film (Kodak).

The proximal E-cadherin promoter contains multiple characterized elements including three E-boxes, a CCAAT box, and a GC-rich element. To assess the role of distinct elements in the regulation of E-cadherin gene transcription, promoter elements were mutated within the context of the reporter gene construct containing E-cadherin promoter sequences extending to −108 of the E-cadherin gene [Ecad(−108)-Luc]. This region was chosen because it is the minimal portion of the E-cadherin promoter demonstrating strong activity in breast cancer cell lines with intact E-cadherin transcription (Ecad+) and greatly reduced activity in breast cancer cell lines defective for E-cadherin transcription (Ecad−; Ref. 2). The three E-box elements in the proximal E-cadherin promoter were mutated, either singly or in combination, from their consensus sequence 5′-CANNTG-3′ in the sense strand to 5′-AANNTA-3′ (Fig. 1,A). Mutations to the outer nucleotides of the 6-bp sequence were selected, because they have been reported previously to abolish factor binding (3). The mutant constructs were then tested in a panel of breast cancer cell lines of known E-cadherin transcription status, and the reporter activity compared with that of the wild-type E-cadherin promoter-driven reporter gene construct Ecad(−108)-Luc. As shown in Fig. 1,B, mutation of the most 3′ E-box, EboxC, resulted in an increase in E-cadherin reporter gene construct activation in Ecad− lines with minimal effect on reporter activity in Ecad+ cell lines. Mutation of all three E-box elements in the proximal E-cadherin promoter additionally increased reporter gene activity in the Ecad− lines, again with minimal effect in Ecad+ lines (Fig. 1 B). These findings suggest the E-box elements negatively regulate E-cadherin transcription in Ecad− lines, potentially via transcriptional repressor(s) binding to one or more of the E-box elements.

Of the three proximal E-box elements, EboxC appeared to play the most significant role in the repression of E-cadherin gene transcription in breast cancer cells (Fig. 1,B; data not shown). Mutation of EboxA alone resulted in an ∼2-fold derepression of E-cadherin reporter gene activity in Ecad− breast cancer cell lines, with no effect on activity in Ecad− breast cancer cell lines (data not shown). EboxB, the central of the three E-box elements, did not appear to significantly modulate E-cadherin gene transcription in either Ecad− or Ecad+ breast cancer cell lines (Fig. 1,C). Our results differ from those reported by other investigators on the relative importance of the E-box elements in the proximal E-cadherin promoter. Specifically, a previous report suggested mutation of EboxB alone resulted in clear derepression of E-cadherin reporter gene activity in carcinoma cell lines (3). Additionally, whereas other studies have also implicated E-box elements in the proximal E-cadherin promoter in regulating E-cadherin transcription in carcinoma cell lines (3, 4, 5), our results offer some new insights. For instance, the most 3′ E-box, EboxC, was the single largest contributor to repression of E-cadherin expression in breast cancer cells (Fig. 1 B).

The three E-box elements in the proximal E-cadherin promoter are of a specific subclass of E-boxes; all contain the sequence 5′-CACCTG-3′. The sequence 5′-CACCT-3′ is known to bind members of the δEF1 zinc-finger transcription factor family, including δEF1/ZEB-1 and SIP1/ZEB-2. Both δEF1/ZEB-1 and SIP1/ZEB-2 have been proposed to repress E-cadherin transcription (5, 7). These factors are characterized by a protein domain structure in which a central homeodomain is flanked by NH₂- and COOH-terminal clusters of zinc-fingers, so that one monomer can bind to bipartite DNA elements (10). The SIP1/ZEB-1 protein has been proposed to repress Ecadherin transcription by simultaneously interacting with both EboxA and EboxB in the proximal promoter (5). Thus, our finding that EboxB is not critical in regulating E-cadherin transcription in breast cancer cell lines (Fig. 1 C) may be of some significance. Given the likelihood that the EboxB mutation we created abolished SIP1/ZEB-1 binding, either SIP1/ZEB-1 is not a critical factor in the repression of E-cadherin in breast cancer or EboxB is not a necessary target for the binding of one zinc-finger of SIP1/ZEB-1 to the E-cadherin promoter.

We next sought to identify and characterize specific proteins that may bind to the E-box elements in the proximal E-cadherin promoter and repress transcription in breast cancer. In light of the established roles of Slug and Snail in the down-regulation of E-cadherin during epithelial-mesenchymal transitions in development (11, 12) and the recent suggestion that SNAIL represses E-cadherin transcription in carcinomas (4), we focused on the role of these factors in the repression of E-cadherin in breast cancer. SLUG and SNAIL belong to the larger Snail family of proteins, and contain an NH₂-terminal repression domain and a COOH-terminal zinc-finger DNA-binding domain (13).

Constructs expressing full-length, flag epitope-tagged SLUG and SNAIL were generated (Fig. 2,A), and the effects of these proteins on E-cadherin reporter gene activity were assessed. When compared with the effects of the empty expression vector on E-cadherin reporter gene activity, both SLUG and SNAIL repressed the wild-type E-cadherin reporter construct in Ecad+ breast cancer cell lines (Fig. 2,B). However, neither SLUG nor SNAIL could repress a construct in which all three of the E-box elements were mutant (Fig. 2,C). Additionally, both SLUG and SNAIL demonstrated dose-dependent repression of the wild-type E-cadherin reporter gene construct Ecad(−108)-Luc in the Ecad+ breast cancer cell line MCF-7 (Fig. 2 D). Taken together, these data show that SLUG and SNAIL are capable of repressing E-cadherin transcription in vitro, and this repression is mediated via the E-box elements in the proximal E-cadherin promoter.

We chose to use a regulatable system to assess the effects of SLUG and SNAIL expression on endogenous E-cadherin expression in epithelial cell lines. For this purpose we used ER chimeric proteins, in which sequences from the protein of interest were fused to a modified ligand-binding domain from the murine ER α. The ligand-binding domain contains a mutation that renders it resistant to binding by endogenous estrogens yet capable of binding to the synthetic ligand 4-OHT (14). Chimeric proteins are constitutively expressed but inactive in the absence of ligand and activated after the exposure of cells to 4-OHT. Constructs expressing SLUG-ER and SNAIL-ER fusions were generated. Stable clones of rat kidney epithelial RK3E cells expressing one of the chimeric proteins were obtained, and expression of the chimeric proteins was confirmed (Fig. 3,A). After 4-OHT treatment for 48 h, E-cadherin transcript levels were decreased in both the SLUG-ER and SNAIL-ER lines, whereas no change in E-cadherin expression was observed in parental RK3E cells (Fig. 3 B).

Given the ability of SLUG and SNAIL to repress endogenous E-cadherin expression in RK3E cells, we sought to study repression of endogenous E-cadherin in E-cadherin-expressing breast cancer cell lines. Attempts to generate stable lines with constitutive expression of SLUG, SNAIL, or stable lines with 4-OHT-regulated SLUG-ER or SNAIL-ER chimeric proteins failed. These data suggest that in the breast cancer cell lines studied, SLUG and SNAIL may have deleterious effects on breast cancer cells when overexpressed. An alternate approach was pursued in which SLUG and SNAIL were introduced into the E-cadherin-expressing breast cancer cell line MDA-MB-468 by retroviral transduction, followed by a brief (2-week) G418 selection to eliminate nontransduced cells. Lysates were then made, and protein expression patterns were assessed. As shown in Fig. 3 C, in this setting, both SLUG and SNAIL decreased E-cadherin protein levels. Thus, both SLUG and SNAIL can repress endogenous E-cadherin in the rat kidney epithelial line RK3E and the human breast cancer cell line MDA-MB-468.

Because both SLUG and SNAIL were capable of repressing E-cadherin promoter activity and endogenous E-cadherin expression, we sought to characterize expression of SLUG and SNAIL in breast cancer cell lines by Northern blot analysis. Expression of SLUG, rather than that of SNAIL, was strongly correlated with the absence of E-cadherin transcripts (Fig. 4). The data imply SLUG is the more likely in vivo repressor of E-cadherin transcription in breast cancer. In cell lines analyzed for their ability to activate E-cadherin reporter gene activities, E-cadherin promoter activities were reduced in lines with SLUG expression (Fig. 1,B; data not shown). However, a tight correlation between the relative levels of endogenous SLUG transcripts and E-cadherin promoter activity was not observed in the three Ecad− cell lines studied, perhaps because SLUG protein levels may not be strictly tied to transcript levels, and differences in the expression of other proteins from one line to another may affect the ability of SLUG to repress E-cadherin transcription. A fraction of the cell lines studied showed seemingly discordant results in regard to SLUG and E-cadherin expression patterns (i.e., SK-BR-3, MDA-MB-468, and BT-20; Fig. 4). Therefore, additional studies were carried out to clarify the relationship between SLUG and E-cadherin in the three lines.

SK-BR-3 has been shown to harbor a homozygous deletion of a large portion of the E-cadherin gene (15). Additionally, it has been demonstrated that SK-BR-3 retains all of the necessary trans-acting factors for E-cadherin transcription and, in fact, activates E-cadherin reporter gene constructs (9). These data are consistent with the view that if a cis genetic mechanism inactivates E-cadherin in SK-BR-3, defects in trans-acting pathways regulating E-cadherin gene transcription would not be expected. The lack of SLUG up-regulation in SK-BR-3 is consistent with this hypothesis. MDA-MB-453 showed E-cadherin and SLUG expression like that seen in SK-BR-3 (Fig. 4). Akin to SK-BR-3, we found that MDA-MB-453 activated E-cadherin reporter gene constructs (data not shown). Thus, the cell line retains the necessary trans-acting factors for E-cadherin gene transcription (Ecad+). A potentially subtle cis genetic alteration at the E-cadherin locus may underlie loss of detectable E-cadherin transcripts, because gross deletions or rearrangements of E-cadherin gene sequences were not seen on Southern blot analysis (data not shown). Alternatively, other mechanisms, such as promoter hypermethylation (16), could contribute to loss of E-cadherin in MDA-MB-453. In the BT-20 cell line both SLUG and E-cadherin transcripts were seen (Fig. 4), suggesting the E-cadherin gene may somehow be resistant to repression by SLUG in this cell line. No mutations were found in the E-box elements or elsewhere in the proximal E-cadherin promoter in the BT-20 cell line (data not shown). Possible explanations for the apparently discordant data are that the SLUG protein is unstable in this cell line, or necessary cofactors for SLUG-mediated repression are not expressed in the BT-20 line. The lack of insights into the identity of such cofactors precluded studies to address this hypothesis.

In closing, we would emphasize that our findings demonstrate the E-box elements, specifically EboxA and EboxC, contained in the proximal E-cadherin promoter appear critical in repression of E-cadherin gene transcription in breast cancer. Whereas SLUG and SNAIL were found to be capable of repressing E-cadherin gene transcription via these E-box elements, our data indicate that SLUG is a more likely in vivo repressor in breast cancer. Given the sizable number of potential transcriptional repressors thus far implicated in the repression of E-cadherin gene transcription (4, 5, 7, 8), it is possible that different factors function in the repression of E-cadherin in different settings. Alternatively, multiple factors may collaborate in mediating repression in vivo. Additional studies should help in resolving the remaining uncertainties about the mechanisms by which SLUG and other factors repress E-cadherin expression in breast and other cancers.

The following investigators provided reagents used in these studies: A. Friedman, Johns Hopkins University, Baltimore, MD; G. Nabel, NIH, Bethesda, MD; G. Nolan, Stanford University, Stanford, CA; and J. M. Rupert, University of Alabama, Birmingham, AL.